Professional Documents
Culture Documents
1.0 Introduction
The Global Positioning System (GPS) offers satellite navigation services that are widely used in
aviation and military applications. The accuracy of received GPS signal and jamming prevention
issues are of utmost importance in designing GPS systems due to its areas of critical applications.
It is important to design GPS systems with efficient anti-jamming capabilities so as to protect the
accuracy of received data. GPS is particularly vulnerable to jamming because the receivers are
extremely sensitive in order to receive the extremely weak signals from orbiting satellites
(Esmaeilkhah and Lavasani, 2018; Lang et al., 2017; Mukhopadhyay et al., 2007).
When the jamming signals and the desired signal share the same frequency, the conventional
filtering techniques become inefficient in eliminating jammers. This has led to the study of
1
Akobundu and Gbenga-Ilori: GPS Anti-Jamming Technique Using smart Antenna Systems AZOJETE, 15(1):1-16.
ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng
various anti-jamming techniques using adaptive antenna system with direction-of-arrival (DOA)
estimation and adaptive beamforming algorithms. These smart antenna systems are made up of
an array of antenna elements with digital signal processors (DSP) to implement the adaptive
signal processing. The objective is to estimate the direction of arrival of all impinging signals and
adjust the antenna weights to ideally steer the maximum radiation of the antenna pattern
toward the desired signal and to place nulls toward the interfering signals. Therefore, adaptive
antenna systems combine multiple antenna elements with a signal processing capability to
optimize its radiation and reception pattern automatically in response to the signal environment
(Joshi and Dhande, 2014). In a smart antenna system, the optimum weights of each antenna
elements are iteratively determined using algorithms based upon different criteria to give a
maximum gain in the direction of the source signal and placing the null in the direction of the
interference signal (Shivapanchakshari and Aravinda, 2017; Shahab et al., 2017).
These algorithms can be categorized into two classes depending on whether a training signal is
used or not. One class of these algorithms is the non-blind adaptive algorithm in which a
training signal is used to adjust the array weight vector (Joshi and Dhande, 2014; Ali et al., 2011).
Another technique is to use blind adaptive algorithms that do not require a training signal (Rao
and Sarma, 2014; Elkassimi et al., 2017; Udawat et al., 2011).
A lot of work has been done on the implementation of anti-jamming systems using non-blind
adaptive beamforming algorithms. Joshi and Dhande (2014) used least mean square (LMS)
algorithm for adaptive beamforming. They evaluated the effect of the LMS algorithm on its
beamforming capability using normalized array factor and the convergence of its mean square
error (MSE). The effect of varying the number of antenna elements and the distance between
array elements on the array factor and MSE was also evaluated. Ali et al. (2011) authors worked
on adaptive beamforming algorithms for anti-jamming. They studied and compared the
performance of three beamforming algorithms; LMS, optimized-LMS and recursive least square
(RLS) algorithms and concluded that RLS algorithm had the best performance as it provided
deeper nulls in the direction of the signal not of interest (SNOI) and faster convergence.
A few others considered blind adaptive algorithms. Rao and Sarma (2014) analyzed and
compared the beamforming capabilities of LMS, Normalized LMS, Constant Modulus Algorithm
(CMA), and RLS for smart antenna application using three different antenna array geometries;
linear, circular and planar geometries. Extensive comparison was done using these algorithms
and geometries. They also compared the beamforming of LMS and NLMS algorithms using
different inter-element spacing and number of elements in the array and concluded that as the
number of elements increased, the algorithms converged faster. Though an extensive analysis
was done for various beamforming algorithms, implementation of complete beamforming
system involving direction of arrival estimation was not considered.
Elkassimi et al. (2017) proposed the use of algorithm based on zero-forcing (ZF) and minimum
mean square error (MMSE) methods for blind channel equalization. They compared performance
of this algorithm with other adaptive filter algorithms like CMA, fractional space CMA (FSCMA)
and sign kurtosis maximization adaptive algorithm (SKMAA). Udawat et al. (2011), the
convergence characteristics of blind beam-forming algorithm known as least square constant
modulus algorithm (LSCMA) were analyzed. The algorithm was used with Smart antenna systems
(SAS) for improving the performance of a wireless communication system by optimizing the
radiation pattern according to the signal environment. Kundu and Ghosh (2008) demonstrated
2
Arid Zone Journal of Engineering, Technology and Environment, March, 2019; Vol. 15(1):1-16. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng
an implementation of real-time adaptive beamforming for code division multiple access 2000
(CDMA2000) reverse link using digital signal processing technique. Simulation was performed
for uniform linear array (ULA) of several elements and they found that algorithm computation
time and accuracy of beam pattern direction were proportional to the number of antenna
elements.
DOA estimation is also a key research area in adaptive array antenna systems. Beulah and
Vigneshwani (2014), focused on the design of a smart antenna system based on multiple signal
classification (MUSIC) DOA estimation algorithm, and adaptive beamforming using LMS. Array
pattern synthesis was achieved using the Dolph-Chebyshev method which is popularly used for
obtaining the weights and current excitation for uniformly spaced linear arrays steered to
broadside. This method produced array pattern with side lobes of equal magnitude and made
use of a class of polynomials called Chebyshev polynomials.
Mukhopadhyay et al., (2007) proposed a new simple DOA estimation method based on the
mechanical rotation of an array plane by small angles. In this method, the frequencies of the
signals impinging on the array must be known. By taking the ratio of these frequencies before
and after rotation of the array plane and comparing with ratio of the cosine function of the angle
of arrival (AOA) before and after rotation, a simple trigonometric equation was obtained for DOA
estimation. Simulations showed that the estimated DOAs and frequencies matched with actual
values. The authors also showed that adaptive beamforming using LMS algorithm can be used
to constructively receive signal with multipath interference.
Very few papers addressed anti-jamming techniques for GPS receivers. In Zhang and Amin
(2012), authors proposed a novel array-based anti-jamming GPS receiver that enables effective
jammer suppression with negligible phase distortion. They pointed out that one of the major
problems with existing blind anti-jamming array signal processing technique is that they
introduce errors in the carrier phase, and developed a technique that achieves GPS signal phase
continuity by using accurate estimation of phase rotated steering vector of the GPS signal. In Chi
(2016), the authors proposed an anti-jamming measurement method based on antenna array.
This method uses array antenna to estimate the direction of arrival of jamming signal through
MUSIC algorithm. The adaptive nulling algorithm based on the linear constrained minimum
variance algorithm is used to eliminate the jamming signals and hold the navigation signal.
According to the above papers, non-blind adaptive algorithms require prior information like the
reference signal and direction of arrival (DOA) of the signal and may not be suitable for GPS
systems. In this paper, the performance of non-blind adaptive algorithm using LMS algorithm
and blind algorithm using CMA algorithm is studied and implemented for adaptive
beamforming while the estimation of signal parameters via rotational invariance technique
(ESPRIT) and MUSIC algorithms have been implemented for DOA estimation purposes. The
purpose is to compare these known techniques and determine the most suitable for an anti-
jamming system in GPS receivers.
4
Arid Zone Journal of Engineering, Technology and Environment, March, 2019; Vol. 15(1):1-16. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng
Figure 2: The functional block diagram of the system implementation (Moghaddam et al., 2011)
By using a cost function, the adaptive algorithm computes the appropriate weights that result in
an optimum radiation pattern. The next subsections will consider in details the steps needed to
implement the algorithms used in the functional blocks.
6
Arid Zone Journal of Engineering, Technology and Environment, March, 2019; Vol. 15(1):1-16. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng
For a ULA, this is achieved by first forming a N × M matrix U which its columns are the M
eigenvectors associated with the largest eigenvalues of the estimated array correlation matrix
Rxx of the full array of N elements. Then select the first L < N rows of U to form Uy and the last of
its L rows to form Uz.
C = Vc ΛVHc ,
where Λ = diag{λ1 , λ2 ,⋯,λ2M} with λ1 ≥ λ2 ≥ … ≥ λ2M are the eigenvalues of C while Vc contains
the eigenvectors corresponding to the eigenvalues.
4. Partition Vc into four matrices of dimension M × M as follows;
V V12
Vc = 11
V21 V22
5. Calculate the eigenvalues λm, m = 1,⋯⋯,M, of the matrix − V11 V−1
22
.
6. Estimate the angle of arrival θm using;
Arg(λm )
θm = sin−1 , m = 1,2, ……, M
2π∆0
____________________________________________________________________________
3. Beamforming Methods
3.1 Beamforming Using LMS
One of the adaptive algorithms that will be used to optimize the weights of the array elements
for adaptive beamforming is the LMS algorithm. Figure 2 shows the output of the system
y(n)and it is given by;
y n = wH n x n . (1)
The error between the desired signal d(n) and system output y(n) is given as;
e n =d n −y n . (2)
The LMS weight update equation is given by;
w n + 1 = w n + μ − ∇J n , (3)
where μ is the step size parameter which controls the speed of convergence and lies between 0
and 1.∇J(n) denotes the instantaneous estimate of the gradient which is given as;
∇ = = 1 . (4)
The gradient estimate can simply be obtained by taking the error value for a single input pattern,
squaring it and differentiating with respect to weight w.
∂e2 n ∂e2 n
(5)
∂e n
∇J n = = ∙ ∂w n ,
∂w n ∂e n
Since e n = d n − WH (n)x(n),
∴ ∇J n = − 2e n x n . (7)
Putting equation (7) into equation (3), the weight update equation is given by;
w n + 1 = w n + 2μx n e∗ n . (8)
Therefore, the LMS algorithm can be described by the three equations as given below;
y n = wH (n)x(n)
e n =d n −y(n) (9)
w n+1 =w n +2μx n e∗(n)
and if u = y n − 1 , then
∂u2 ∂u2 ∂u
∇J w = E ∂w
=E ∙
∂u ∂w
(12)
1 1
∂(y(n)y∗(n)) 2 ∂(w n H x(n)x n H w(n)) 2
= E 2u ∙ ∂w
= E 2u ∙ ∂w
(13)
If w n H x n x n H w n = v, then
1 1
(14)
∂v 2 ∂v 2 ∂v
∇J w = E 2u ∙ ∂w
= E 2u ∙ ∂v
∙
∂w
−1
(15)
v 2 y n x(n)
= E 2u ∙ ∙ x n x n H w(n) = E 2( y n − 1) ∙
2 2|y(n)|
8
Arid Zone Journal of Engineering, Technology and Environment, March, 2019; Vol. 15(1):1-16. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng
Removing the expectation parameter gives the instantaneous gradient value as;
(16)
y n
∇J w = y n − y n
x n
Equation (16) can then be used to replace the gradient term in the traditional LMS weight
update algorithm of equation(3) to yield;
(17)
y n
w n + 1 = w n − μx n y n − y n
This is similar to LMS but with update error, e, expressed as;
(18)
y n
e n =y n − y n
Therefore the CMA can be described by the three equations set given below;
y n = wH (n)x(n
y(n)
e n = y n − |y(n)| (19)
w n + 1 = w n − μx n e∗ (n)
Figure 5 shows the MUSIC and ESPRIT spectrum of the estimated angles of arrival of both the
desired and interference signals. From the figures, it can be seen that both DOA algorithms gave
an accurate estimate of the directions of arrival. Since both DOA algorithms were able to
estimate the correct angles of arrival, only one of the algorithms was used for other results as
they produced a similar response.
The plots of Figure 6 show the radiation pattern and amplitude response of the array output in
the simulations. From the plots, it can be seen that both beamforming algorithms used were
able to give maximum radiation in the direction of the desired GPS signal ( 0° ). In terms of
placing nulls in the directions of interference, LMS algorithm had the best response by placing
deeper nulls in these directions (50° and − 60°) whereas CM algorithm had a shallow null at 50°
and couldn’t place a null at − 60° though a minor side lobe was formed. This is because CMA is
less stable and presented some errors during implementation and as shown in subsequent
results. The figure shows that the algorithms achieved the goal of the system by placing
maximum radiation in the direction of the desired signal.
The plots of Figure 6 shows the radiation pattern and amplitude response of the array output in
the simulations. From the plots, it can be seen that both beamforming algorithms used were
able to give maximum radiation in the direction of desired GPS signal ( 0° ). In terms of placing
nulls in the directions of interference, LMS algorithm had the best response by placing deeper
nulls in these directions ( 50° and − 60° ) whereas CM algorithm had a shallow null at 50° and
couldn’t place a null at − 60° though a minor side lobe was formed. This is because CMA is less
stable and presented some errors during implementation and as shown in subsequent results.
10
Arid Zone Journal of Engineering, Technology and Environment, March, 2019; Vol. 15(1):1-16. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng
The figure shows that the algorithms achieved the goal of the system by placing maximum
radiation in the direction of the desired signal.
0
0
-10
-10
-20
-20
-30
-30
-40
-40
|P( )|
-50
|P( )|
-50
-60
-60
-70
-70
-80
-80
-90 -90
-100 -100
-100 -50 0 50 100 150 -100 -50 0 50 100 150
Angle Angle
(a) (b)
Figure 5: (a) MUSIC spectrum of estimated angles of arrival of the signals (b) Equivalent ESPRIT
spectrum.
(a) (b)
Figure 6: (a) Radiation pattern of the system after beamforming (b) Polar representation
Figure 7 depicts the Mean Square Error (MSE) plot or convergence plot for the CMA and LMS
algorithm respectively. It was observed that the LMS algorithm converged to a near-zero value
before 100 iterations whereas implementing the system with CMA presented lots of errors. It
was also observed that the CMA attempted to converge at some point, but due to the discrete
nature of the desired signal as shown later, erroneous results were obtained while tracking of the
amplitude levels is done.
Figure 8 shows the acquisition and tracking curve of the system. It can be seen that the system is
excellent in tracking the reference signal when the LMS algorithm is implemented and there is
no delay in the output of the system to align perfectly with the reference signal. On the other
hand, it was difficult for the output of the system to track and align to the reference signal when
the CM algorithm is implemented. Finally, a phase comparison of the system output and the
reference signal as shown in figure 9 indicates that there is no much difference between the two
when either of the algorithms is used.
11
4
Mean square error
Magnitude
3
2
0.5
0 0
0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
Simulation iteration Samples
-100
Phase(rad)
-150
-200
-250
-300
0 100 200 300 400 500 600 700 800
Samples
Figure 9: Phase comparison of the system output and the reference signal
12
Arid Zone Journal of Engineering, Technology and Environment, March, 2019; Vol. 15(1):1-16. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng
(a)
(b)
Figure 10: Effect of inter-element spacing on beamforming (a) d = λ, (b) d = λ 8
In Figure 12, the number of elementsN is varied between 8, 20 and 50, and it was observed that
as the number of elements in the array increases, estimated DOA spectral beamwidth becomes
narrower, the directivity of the array is enhanced meaning that its ability to distinguish spatial
signals is improved. Therefore an accurate DOA estimation can be achieved by increasing the
13
(a)
(b)
Figure 11: Effect of number of elements N on beam pattern (a) N = 20, (b) N = 50
Figure 12: Effect of number of elements N on DOA estimation using MUSIC spectrum
14
Arid Zone Journal of Engineering, Technology and Environment, March, 2019; Vol. 15(1):1-16. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng
5. Conclusion
This paper focused on the implementation of an anti-jamming system using DOA estimation and
adaptive beamforming algorithms for GPS application. MUSIC and ESPRIT DOA estimation
algorithms, as well as LMS and CMA adaptive beamforming algorithms, were considered in this
study. MATLAB was used to develop these algorithms and to model the desired (GPS) and
interference signals. Simulation results presented showed that for the DOA estimation, both
algorithms are high-resolution algorithms as seen by the accuracy of the estimated direction of
arrival although MUSIC algorithm produced a better direction of arrival spectrum with little or no
minor peaks. Also for the beamforming, both LMS and CMA produced maximum radiation in the
direction of the desired signal. LMS placed deeper nulls in the directions of interference with
faster convergence and fewer errors as compared with CMA, which presented errors and was
able to suppress the interference to an extent. It was also shown that as the number of elements
in the array increases, a more directive beam and DOA spectrum is produced. In conclusion, the
results show that smart antenna system is viable and very instrumental in the development of
anti-jamming systems.
References
Ali, RL., Ali, A., Khan, SA. and Malik, SA., 2011. Adaptive beamforming algorithms for anti-
jamming. International Journal of Signal Processing, Image Processing and Pattern
Recognition, 4(1): pp.95-106.
Beulah, RE. and Vigneshwari, NA., 2014. Design and Simulation of Smart Antenna Array Using
Adaptive Beam forming Method. IJESIT, Vol. 3, Issue 6: pp. 434-441.
Borre, K., Akos, DM., Bertelsen, N., Rinder, P. and Jensen, SH., 2007. A software-defined GPS and
Galileo receiver: a single-frequency approach. Springer Science & Business Media.
Elkassimi, S., Safi, S. and Manaut, B., 2017. Blind Channel Equalization by Adaptive Filter
Algorithms. Universal Journal of Applied Mathematics, 5(3): pp.45-52.
Esmaeilkhah, A. and Lavasani, N., 2018. Jamming Efficacy of Variable Altitude GPS Jammer
against Airborne GPS Receiver, Theoretical Study and Parametric Simulation. Advanced
Electromagnetics, 7(1): pp.57-64.
Gao, F. and Gershman, AB., 2005. A generalized ESPRIT approach to direction-of-arrival
estimation. IEEE signal processing letters, 12(3): pp.254-257.
Huang, L., Wu, S. and Zhang, L., 2005, January. A novel MUSIC algorithm for direction-of-arrival
estimation without the estimate of covariance matrix and its Eigen decomposition. In Vehicular
Technology Conference, 2005. VTC 2005-Spring. 2005 IEEE, Vol. 1: pp. 16-19.
Joshi, R. and Dhande, AK., 2014. Adaptive beamforming using LMS algorithm. International
Journal of Research in Engineering and Technology, 3: pp. 589-593.
Kintz, AL. and Gupta, I.J., 2016. A modified MUSIC algorithm for direction of arrival estimation in
the presence of antenna array manifold mismatch. IEEE Transactions on Antennas and
Propagation, 64(11): pp. 4836-4847.
15
16